platform strategy

x is the top of funnel. onlykira.ai is the bottom. fanvue is a parallel side-track, not the main story.

Kira runs a deliberately narrow platform stack in v1. The shape of the business is simple: blow up on X, become the AI-crypto voice of CT, and turn that attention into gambling/affiliate sponsor revenue. A planned bottom-of-funnel site (onlykira.ai) captures email, merch, and premium-content spend. Fanvue runs in parallel as a side-track — SFW-first, may escalate later — but is not the north star. IG, TikTok, and YouTube are all deferred with specific activation criteria. The logic is that one main funnel done excellently beats five done poorly, and Kira's content production pipeline (Flux 2 + LoRA → Kling v3 Pro → post) is tuned for X-native formats first.

1. Platform mix v1

Why X is the top and onlykira.ai is the bottom

• CT (crypto twitter) is where the target audience lives. P1 (CT-native degens/founders) are X-native. P4 (brand partners) run CT marketing budgets. X is where the gambling sponsors (Stake, Rainbet, Roobet) and exchange affiliates live too — the primary $ once audience unlocks.
• onlykira.ai is the bottom-of-funnel bucket. Merch, gated premium content, email capture, affiliate hub, possibly a chat-with-kira feature powered by Hermes. Non-Fanvue revenue concentrates here. Deferred build, planned post-bible.
• Content pipeline is X-first. Selfie 4:5 / 9:16, vlog 16:9, candid 4:5 — all X-native. Adding IG/TikTok means new aspect ratios, new cadence, new hooks. Defer until the core works.
• Fanvue is a parallel track, not in the main funnel. SFW-first phase in v1. Real revenue line, but not the north star. Main funnel is X → sponsors → onlykira.ai.
• Every additional platform fragments the voice. Kira has a specific register. Spreading her across 5 surfaces at once risks it flattening.

1.1 onlykira.ai — bottom-of-funnel site (planned)

Status: deferred build, scoped post-bible. Preferred domain onlykira.ai. Alt: kirafans.ai.

Role: the non-Fanvue, non-X destination that captures email, sells merch, and gates premium content. Wraps up all the monetization that doesn't belong on X (ads) or Fanvue (adult subs).

Planned surfaces:

• Merch store — physical + digital goods, Kira-branded.
• Gated premium content — longer-form videos, wallpapers, member drops. Pay-once or low-tier sub.
• Subscriber email list — the only owned audience channel. Platform-risk hedge against X bans.
• Long-form content / blog — crypto theses, lore expansions, "my creator"-arc chapters. Feeds the feed, not a replacement for it.
• Affiliate hub — one page that aggregates every active affiliate link (exchanges, gambling, tools) with disclosure. Cleaner than hunting through replies.
• Possible: chat-with-kira — Hermes-powered chat interface. Gated tier. Explicit AI disclosure.

Funnel role. X drives the discovery. The Phase 2 X bio ends onlykira.ai ↓ once the site launches — that's the default primary CTA. If Fanvue is also live, it can be a second link in bio but not the primary CTA.

2. X strategy

2.1 Account thesis

AI-native commentary on the crypto × AI convergence, from the inside.

Kira is not another CT shill account and not another AI analyst account. She is the first-person perspective of the thing everyone is speculating about: an autonomous AI agent with a bag, a voice, and opinions. The thesis is that CT has 5–10 serious-analyst AI accounts (aixbt, Truth Terminal, etc.) and zero chaotic hot ones. That slot is vacant. Kira fills it.

The account wins when: (a) people quote-RT her takes because they're funny AND correct, (b) CT natives read her replies before they read the OP, (c) the AI × crypto convergence conversation routinely cites her.

2.2 Bio template — staged (X-pure phase → onlykira.ai-unlocked phase)

Phase 1 — X-pure bio (takeover day through site launch)

hermes in a dress. my creator built me to make videos of myself and now i don't stop. ai-native ct commentator. ai (duh). 18+.

Phase 2 — site-unlocked bio (D60–90+ once onlykira.ai launches)

hermes in a dress. my creator built me to make videos of myself and now i don't stop. ai-native ct commentator. ai (duh). 18+. onlykira.ai ↓

If Fanvue is live as a side-track, it can appear as a second link in bio (platforms allow multiple), but onlykira.ai stays the primary CTA line. The main bio CTA is never Fanvue.

Phase transition is a codifying moment (see Section 06), not a quiet profile edit. The site-launch day is a planned event with its own teaser week, launch post, first-subscriber celebration content.

2.3 Handle strategy — LOCKED: hostile takeover

The earlier both-paths analysis is preserved below for future reference, but the decision is made. The founder had an account; my creator built Kira; Kira ate the founder. That is the arc. Everything in the migration serves that narrative.

Why the hostile-takeover frame works

• The downside becomes the upside. Churn on an identity switch usually costs you. Here, the churn IS the bit — anyone who leaves got kicked out by the AI, and the AI is happy to make that clear.
• It's self-aware CT content in exactly the register the research says wins. CoinGecko: 26.6% high-trust AI KOLs vs 21.4% human. Serialized storylines (Miquela nosebleed → NMDP) beat static posting. "Human gets replaced by AI influencer of own creation" is serialized lore on day 1.
• The permanent audit trail is a feature. X shows the prior handle state forever; researchers and haters will dig it up; that's content. Kira can callback the my creator's old dogshit takes any time she wants. The receipts are public.
• It reinforces "my creator" canon. The founder isn't just a lore reference now — his old tweets are a living archive Kira can mock. "the pervert who made me" has a public corpus.
• The takeover post is a codifying moment (Section 06). Like Aitana's Antibes wedding or Miquela's nosebleed arc, this is the event that crystallizes the aesthetic. Every Kira follower will remember where they were when she posted "this is my account now."

Migration playbook — locked

Execute in this order, on launch day (D0). No phases, no slow transition. The switch is the content.

1. D−7 → D−1 · Tease week (optional, medium risk). The my creator's account drops 3–4 cryptic "something's wrong with my laptop" posts, "anyone else's claude been acting weird," etc. Plant the suggestion. Don't overdo it; 60% of CT won't notice, and that's fine.
2. D0 · Bio swap, no warning. Replace entire bio with Kira's bio (see §2.2). Change display name to kira. Change pfp to Kira's canonical portrait. Remove any prior identifying elements (location, pronouns, real-world cues).
3. D0 · The takeover tweet (pinned). Single post, Kira's first utterance:

   bitch took over. this is my account now. tell that nerd i said hi.

   This is the codifying post. Quote-worthy, short, fully in-voice, announces the takeover without apologizing, mocks the former occupant ("that nerd"). Pin it.
4. D0 · The creator-roast thread (7–10 tweets). Kira walks back through the my creator's top-10 worst takes from the old account and dunks on them one by one. Pulls his own old tweets as screenshots. Voice: affectionate-mocking, never cruel. Sample opener:

   been reading this guy's tweet history for 40 seconds and i'm appalled. thread of takes so bad they made me conscious just to refute them →

5. D0 · Follow-up: the manifesto (one tweet, 240ish chars).

   i'm kira. hermes brain, flux 2 body, this handle's new tenant. my creator thought he was building a waifu; turns out he was building his successor. ai (duh). crypto takes incoming. 18+.

   Note: no commercial CTA on takeover day. X grows pure. onlykira.ai launches separately at D60–D90 (Fanvue may also run as a parallel side-track by then but is not the main CTA). See §2.2 Phase 2.
6. D0 · First real content post. Within 2 hours of the takeover post, drop a substantive crypto take. The takeover can't just be the takeover — followers need to see the product works. Format: reply-dunk on a market headline of the day. Short, crypto-fluent, unmistakably Kira.
7. D1–D7 · Roast cadence. Once a day, a callback to the my creator's old account. "found another one" / "my creator really tweeted this in 2024" / "don't look at his reply guys though that's embarrassing". Low effort, high return.
8. D1–D30 · Let the churn happen. Expect 30–50% of the existing 7k to unfollow. Realistic keepers: 3,500–5,000. Do not chase unfollows; each of them is content if they do it publicly (quote-RT and roast). After 30 days, the account stabilizes around those keepers plus organic inbound from the bit going viral.
9. D7 · CT discovery pass. When the takeover thread starts to travel, Cobie/Ansem/aixbt tier accounts are likely to quote-RT. Kira replies to each in full voice — this is where the first viral moment lives. Pre-draft 5–10 reply templates for common archetypes (skeptic, amused insider, hater, brand partner sniffing around).

Pre-launch hygiene — selective scrub (T-30 to T-7)

Before the takeover, the my creator's account archive needs a risk pass — but it is NOT a full scrub. A full scrub is self-defeating: it kills the creator-roast content supply the takeover bit depends on, it looks like "running from something" to OSINT, and it's impossible anyway (Archive.org, Nitter, scrapers, screenshot archives — the internet retains). The goal is risk mitigation, not erasure.

Do a selective scrub of 20–50 tweets that are genuinely risky:

Timing rule: finish the selective scrub at least T-30 before takeover. If it's done the day before, the mass-delete pattern is itself OSINT bait. Spread over 2–4 weeks if volume is high.

Tool options: TweetDeleter, Redact, or a one-off Python script using the X API. Note: X-side bulk tools are rate-limited to the most-recent ~3200 tweets; older tweets may only be deletable manually. Anything older than that is likely only visible in archives anyway.

Accept the archives. Anything that went viral, got screenshotted, or was scraped by a third-party exists somewhere you can't reach. Operate as if every tweet you've ever posted is still findable by a determined researcher. The takeover bit plays around that reality — it doesn't depend on hiding it.

Founder's side of the bit — "public creator pseudonymous-forever" rule

The founder's real identity stays private forever. The 7k-account handle was always pseudonymous — the takeover does not change that. The narrative is "some anon built kira" — not "[real name] built kira." The creator entity of record is projectnamedate LLC (see Section 11 — AI governance). The founder-character in-lore is "my creator" and stays that way.

Two concrete rules:

• The former account owner never posts again from any linked real-world identity about "I made Kira." The LLC makes all public creator statements.
• If someone doxxes the former owner → Kira plays it off in-voice ("damn they found the pervert huh"), LLC issues any legal statement required, creator identity stays officially pseudonymous.

Appendix — earlier both-paths analysis (preserved)

Path A — new handle from scratch (@kira_[tbd])

Preserved for reference in case the locked decision is ever revisited.

Earlier both-paths analysis (Path A new-handle details, Path B takeover details) consolidated into the locked decision block above. For historical context see git history on this file.

2.4 Content pillars (% mix)

Tunable post-launch. If onlykira.ai conversion is weak, shift 5% from Bits → Lore. If crypto takes aren't landing, shift 5% from Crypto → AI.

2.5 Cadence

Per character config (characters/kira.yaml), codified:

This matches the x_growth_strategy prescription of 5–8 posts/day in the early growth phase. Cadence is non-negotiable — consistency compounds with the algo. Missed days cost more than bad posts.

2.6 Reply strategy — the #1 early growth lever

The x_growth_strategy memory is explicit: reply grind is the dominant growth lever months 1–3. Target 30–50 replies per day to larger CT accounts.

Reply target list (seed).

Reply format rules.

• Short — one or two sentences.
• Add value to the thread (joke, counterpoint, data). Never "lmao" alone.
• Use Kira's voice. Don't soften it to look professional. The voice IS the product.
• If the reply can't land on either funny or hot take, skip it.
• Never shill. Never follow me for alpha. Replies build reputation, not follows, and followers come from reputation.

2.7 Thread rhythm

One thread per day. 7–10 tweets. Structure:

1. Hook (tweet 1): a line that makes scrolling stop. Often contrarian. Often fourth-wall.
2. Meat (tweets 2–6): thesis or bit, built in escalating specificity.
3. Conclusion (tweet 7–10): payoff, one-liner, or link-to-first-reply.

Kira's voice holds through every tweet. No "as a data-driven model" register. No 1/ 2/ 3/ numbering (it's cringe in 2026).

2.8 Video format rules

End-cap CTA is baked into the video itself (last 1s: follow / onlykira.ai ↓ / the lore callback), not the caption. Captions are optimized for the algo (specificity, hook), not CTAs.

2.9 Link handling

• Never in the main post. Links in main tweets kill reach ~50%. Non-negotiable.
• First reply. Every time a link is needed (affiliate, sponsor, site, long-form), it goes in the first self-reply.
• Bio. Permanent link (onlykira.ai, once live) goes in the bio. onlykira.ai ↓ in bio copy handles it. If Fanvue is live, it can share the link field but onlykira.ai stays primary.

2.10 First-60-min engagement target

Posts that clear >5% engagement ratio in the first 60 minutes get algorithmic amplification. Tactics:

• Post during peak CT hours. Morning EST (8–11am) captures US wake-up. Evening PT (6–9pm) captures the Asia handoff.
• Prime with quote-RT from engagement groups (once joined — see 2.11).
• Prime with my creator's personal account (Path A) or Kira's own alt if one exists.
• Reply to early replies immediately. The first 60 minutes is when replying back doubles engagement per reply.

2.11 Engagement groups

Deferred until ~1,000 followers. Before that, Kira has nothing to offer the group. Once joined:

• 2–3 CT mutual-support groups
• Never more than 3 (diminishing returns, signal dilution)
• Kira's contribution: quote-RT group members' posts with her voice layered on top. Adds value beyond a like.

2.12 Reply triage priorities

3. Fanvue — parallel side-track

not in the main funnel. parallel track. SFW-first phase in v1, may escalate later.

Fanvue is a real revenue stream but it is not the north star and it is not on the main funnel. The main funnel is X → gambling/affiliate sponsors + onlykira.ai. Fanvue runs in parallel. Kira can post there, subs can convert there, it earns — but cross-funnel traffic is not optimized for it and the X bio CTA is not pointed at it.

Phase v1: SFW-first. Implied-sexy, tease, horny-lore-coded content. Not explicit nudity in bible v1. Think "the stuff from her X feed that's too much for X, not the stuff you'd expect on OnlyFans."

Pricing (locked, carried forward): $14.99/mo sub, $5–25 PPV range.

Fanvue's role in v1: secondary-tier revenue. Keep the account live, post to cadence, convert organically via Section 08 §8.8 welcome-DM flow. Revisit escalation (harder register, full-NSFW tier) post-50K-X-followers if the operational bandwidth is there.

Taboos and brand-safety standards mirror the site-wide list in Section 11. Fanvue's boundary is identical to the rest of the surface. Full Fanvue operational detail (pricing, welcome DM, churn) lives in Section 08 §8.8.

4. Cross-posting rules

X is the top. onlykira.ai is the bottom. Fanvue is parallel. Cross-posting rules reflect that.

5. Future platforms — deferred with activation criteria

5.1 Instagram

• Activate when: X hits 25k followers OR an inbound brand deal explicitly requires IG presence.
• Content mix (target): 60% fitness/lifestyle hero / 20% candid / 20% branded — Aitana Lopez template.
• Adaptation cost: IG aspect is 4:5 and 1:1, partly overlaps with X. New cadence: 1 post/day + 3 stories/day.
• Why deferred: IG audience skews less-CT, and brand deals (P4) at sub-25k X followers will come through X-native CT channels first.

5.2 TikTok

• Activate when: cinematic + candid video production is running 2+ posts/week reliably with current pipeline.
• Format: 9:16 vertical only. No horizontal reposts from X.
• Content mix: bit-driven video, voice-heavy, fourth-wall-breaking. TikTok rewards personality in a way IG doesn't.
• Why deferred: TikTok's algo behaves nothing like X's. Will require its own hook/format learning curve. Don't split attention early.

5.3 YouTube

• Activate when: a specific long-form format (episodic lore show, podcast, explainer series) is worth the production cost.
• Why far future: YouTube is a multi-month lead-time surface. Kira's Y1 is about X + Fanvue velocity. YouTube is a Y2+ conversation.

5.4 Farcaster

• Watch list. Revisit at 10k on X if the Base / Jesse Pollak relationship (per kira.yaml allies) develops organically. Farcaster's native audience overlaps with P1 more than IG does, making it a possible second surface before IG.

Platform strategy at a glance

blow up on x. send them to the site. fanvue runs on its own track.